This subproject is one of many research subprojects utilizing theresources provided by a Center grant funded by NIH/NCRR. The subproject andinvestigator (PI) may have received primary funding from another NIH source,and thus could be represented in other CRISP entries. The institution listed isfor the Center, which is not necessarily the institution for the investigator.The main objective of the BioWar Project is to develop computational capability for generating realistic simulations in sufficient details of weaponized disease spread on a realistic population. This project is of great importance to policy analysts, policy makers, health officials, first responders, and common folks who need a reasonably accurate tool to think about, plan for, and respond to the event of disease agent release. In addition to bioterrorism, the project is of great importance to public health in general. The recent scare of Severe Acute Respiratory Syndrome (SARS) accentuates the importance and the urgency. BioWar, the large-scale agent-based social-network modeling of how weaponized diseases spread in a population, has run on several Pittsburgh Supercomputing Center's systems, including Jonas. The BioWar simulations typically run for 5 hours for 100,000 agents for 2 year simulation-time period. The last stable version of BioWar code is multithreaded; we just recently compiled the code without threading for Cray XT3's Catamount operating system. We had only a short time access to Lemieux (of PSC), so a parallel version of BioWar is not developed. Moreover, validation of BioWar simulations is needed to endow the basically social and epidemiological simulations with sufficient confidence to be used like how weather simulations are used and trusted to forecast incoming hurricanes. We are developing an automated validation tool called Wizer for this purpose. BioWar is built upon finite state machines, the Construct cognitive-structural co-evolution algorithm, and social networks. Unlike traditional models which use stylized cities, BioWar loads real city demographics, census data, geographical boundaries, weather, school districts, etc. And unlike conventional epidemiological models, BioWar allows tracking of secondary outcomes such as school absenteeism and drug purchase, in addition to infection and mortality rates. BioWar has simulated San Diego, San Francisco, DC, Norfolk, Hampton, and Pittsburgh. Wizer is built upon production systems enhanced with causal reasoning, statistical tests (e.g. bound checking), and ontological reasoning. Wizer takes the inputs of simulation outcome, simulator causal diagrams, simulation parameters, and empirical data, and performs reasoning about which parameter and meta-model variable should be changed to produce a better fit between the simulation and the empirical data. We have done manual validation of BioWar against conventional epidemiological models of anthrax and smallpox. What Wizer adds is the automation of validation, which assists human experts in reasoning about simulation outcomes and in suggesting what the next simulation should be. The purpose for this Development Allocations Committee (DAC) or start-up request of Cray XT3 'BigBen' at the Pittsburgh Supercomputing Center is six folds: (1) running non-threaded single-processor BioWar code on multiple-nodes to gather statistics for model validation, (2) developing automated validation tool (Wizer) for a single processor, taking advantage of multiple-node BioWar runs, (3) developing a parallel version of BioWar, (4) developing a parallel version of Wizer, (5) preparation for submitting a Medium Resource Allocations Committee (MRAC) proposal for BioWar and Wizer, and (6) training and familiarization of Cray XT3 development environment, including but not limited to pgCC compiler, MPI-2 message passing library, Apprentice and Craypat performance tools, and Totalview debugger. The BioWar project was supported by DARPA's Disease Surveillance grant and the Army Research Labs.
Showing the most recent 10 out of 292 publications
